Supervisors info:
Θωμαΐδης Νικόλαος- Αναπλ. Καθηγητής ΕΚΠΑ, Κουππάρης Μιχαήλ - Καθηγητής ΕΚΠΑ, Ευσταθίου Κωνσταντίνος - Καθηγητής ΕΚΠΑ
Summary:
Contaminants of emerging concern enter daily into the sewage system in large
quantities. Concentration levels of the different compounds are not constant,
and may follow different trends and patterns, affected by several factors (like
usage trends, or pollution spills). The investigation of the concentration
pattern of both target and non-target substances in wastewater is an important
issue, as it could be used as an early-warning system on pollution loads, it
may provide a further insight on the behavior of organic contaminants in the
environment, and also information about the community use of chemicals.
The objective of the presented study is to develop an semi-automated workflow
for the detection of trends in intensities for all the detected substances
among different sampling sets (e.g. time periods) through the use of a workflow
based on R algorithms from the packages XCMS, CAMERA and TIMECOURSE. The output
of this workflow is a list of compounds with high variation during the sampling
period extracts. The identity of those compounds was searched following non
target procedures and techniques.
Temporal sequences of samples were collected from the wastewater treatment
plant of Athens. Analysis was carried out by liquid chromatography - high
resolution tandem mass spectrometry (LC-HRMS). First, the acquired data is
converted to mzXML using ProteoWizard and stored in subfolders in the R working
folder. Sample feature detection is performed by the centWave algorithm with
optimized parameters for QTOF MS data. Features representing the same analyte
across samples are placed into groups, peak alignment follows, missing features
are filled with a low intensity value and then, features are clustered
according to peak shape correlation coefficients and retention times. Finally,
isotopic peaks and adducts are annotated to the same chemical component and its
monoisotopic peak. The peak table with the integrated peak areas in each sample
is used in order to perform Multivariate Empirical Bayes Approach which results
in providing prioritized peaks (using Hotelling T2 coefficient) according to
the differences of integrated intensities among the studied groups. Moreover,
it automatically produces plots to evaluate the trend of each substance. After
that, the most relevant peaks can be tentatively explored using non-target
identification approaches.
The presented approach enables the recording of concentration trends for a
large number of compounds in a given set of samples. Moreover, this workflow
can be used to detect events of direct disposal of some specific substances
into the sewage system, constituting an appropriate source of information for
WWTP authorities.
Keywords:
Trends analysis, Statistical language R, Emerging contaminants, Non-target screening, Wastewater
